microRNA expression profiling identifies a four microRNA signature as ...

5 downloads 14481 Views 2MB Size Report
signature as a novel diagnostic and prognostic biomarker in triple negative ..... IHC or high/low miRNA signature status (data not shown) we proceeded with the ...
Oncotarget, Vol. 5, No. 5

www.impactjournals.com/oncotarget/

microRNA expression profiling identifies a four microRNA signature as a novel diagnostic and prognostic biomarker in triple negative breast cancers Pierluigi Gasparini1,*, Luciano Cascione1,8,*, Matteo Fassan1,2,*, Francesca Lovat1, Gulnur Guler3, Serdar Balci4, Cigdem Irkkan5, Carl Morrison6, Carlo M. Croce1, Charles L. Shapiro7, Kay Huebner1 1

Department of Molecular Virology, Immunology and Medical Genetics, Ohio State University Wexner Medical Center and Comprehensive Cancer Center, Columbus, Ohio, USA 2

ARC-NET Research Centre, University and Hospital Trust of Verona, Verona, Italy

3

Department of Pathology, Hacettepe University, Ankara,Turkey

4

Department of Pathology, Yildirim Beyazit University, Ankara Ataturk Research and Training Hospital, Ankara, Turkey

5

Dr. Abdurrahman Yurtaslan Ankara Oncology Training and Research Hospital of the Ministry of Health

6

Department of Pathology, Roswell Park Cancer Institute, S606 Basic Science Building, Elm and Carlton Streets Buffalo, NY

7

Division of Medical Oncology and the Breast Program James Cancer Hospital and Ohio State University Comprehensive Cancer Center, Columbus, Ohio, USA 8 *

IOR,InstituteofOncologyResearch,Bellinzona,Switzerland

These authors contributed equally to the work

Correspondence to: Pierluigi Gasparini, email: [email protected] Keywords: Triple Negative breast cancer, microRNA, five markers, prognosis, treatments, outcome Received: December 11, 2013

Accepted: January 19, 2014

Published: January 21, 2014

This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

ABSTRACT: Triple Negative Breast Cancers (TNBC) is a heterogeneous disease at the molecular and clinical level with poor outcome. Molecular subclassification of TNBCs is essential for optimal use of current therapies and for development of new drugs. microRNAs (miRNA) are widely recognized as key players in cancer progression and drug resistance; investigation of their involvement in a TNBC cohort may reveal biomarkers for diagnosis and prognosis of TNBC. Here we stratified a large TNBC cohort into Core Basal (CB, EGFR and/or CK5, 6 positive) and five negative (5NP) if all markers are negative. We determined the complete miRNA expression profile and found a subset of miRNAs specifically deregulated in the two subclasses. We identified a 4-miRNA signature given by miR-155, miR-493, miR-30e and miR27a expression levels, that allowed subdivision of TNBCs not only into CB and 5NP subgroups (sensitivity 0.75 and specificity 0.56; AUC=0.74) but also into high risk and low risk groups. We tested the diagnostic and prognostic performances of both the 5 IHC marker panel and the 4-miRNA expression signatures, which clearly identify worse outcome patients in the treated and untreated subcohorts. Both signatures have diagnostic and prognostic value, predicting outcomes of patient treatment with the two most commonly used chemotherapy regimens in TNBC: anthracycline or anthracycline plus taxanes. Further investigation of the patients’ overall survival treated with these regimens show that regardless of IHC group subdivision, taxanes addition did not benefit patients, possibly due to miRNA driven taxanes resistance. TNBC subclassification based on the 5 IHC markers and on the miR-155, miR-493, miR-30e, miR-27a expression levels are powerful diagnostic tools. Treatment choice and new drug development should consider this new subtyping and miRNA expression signature in planning low toxicity, maximum efficacy therapies. www.impactjournals.com/oncotarget

1174

Oncotarget

INTRODUCTION

influencing diverse biological processes [10, 11] through effects of individual miRNAs on translation of multiple mRNAs. We previously determined miRNA expression profiles and expression profiles of a cancer-focused mRNA panel, in breast cancers, adjacent non-tumor (normal) and lymph node metastatic lesion (mets) tissues, from 173 women with TNBCs; we then linked specific miRNA signatures to patient survival and used miRNA/mRNA anti-correlations to identify clinically and genetically different TNBC subclasses. For the current study we stratified the TNBC cohort based on EGFR and CK5/6 scoring, into CB and 5NP subgroups and determined if there is a subset of miRNAs specifically deregulated in one or the other of the two subclasses. We indeed found a four miRNA signature which allowed subdivision of TNBCs into two subgroups (high risk and low risk) for which associations among specific clinical features and the four miRNAs were then sought, including outcomes based on specific chemotherapy regimens. The role of specific chemotherapy agents in the treatment of TNBC remains incompletely defined. Taxanes and anthracyclines are active in TNBC and remain important agents, but have not shown specific benefit for TNBC patients vs non-TNBC [12, 13]. Although TNBC is associated with a poor prognosis, some patients respond well to anthracycline-based chemotherapy, reflecting a significant degree of molecular heterogeneity within this subgroup [14-16]. Thus, we have examined and compared the prognostic value of the IHC based subclassification in CB and 5NP and of a four miRNA signature efficacy relative to specific chemotherapy regimens.

Triple-negative breast cancers (TNBC), defined by the absence of estrogen receptor, progesterone receptor, and HER-2 expression, account for 12% to 24% of all breast cancers. TNBCs are associated with early recurrence of disease and poor outcome. Through gene expression profiling, six different TNBC subtypes, defined by abrogation of signaling pathways, have been identified: basal-like 1 and 2 (BL1 and BL2), immunomodulatory, mesenchymal, mesenchymal stem-like, and luminal androgen receptorexpressing [1]. These molecular entities have shown significant differences in terms of incidence, risk factors, prognosis and response to treatment [1-3]. Approximately 15% of breast cancers are basallike and are associated with poor relapse-free and overall survival [4-6]. The basal-like subtype is of particular clinical interest due to its high frequency, lack of effective targeted therapies, poor baseline prognosis, and tendency to affect younger women. Over the years, basal-like breast cancer has become commonly known as triple-negative breast cancer (TNBC), lacking estrogen receptor (ER) and progesterone receptor (PR) expression as well as human epidermal growth factor receptor 2 (HER2) amplification; however, not all TNBCs are identified as basal-like by gene expression, and not all basal-like tumors are Triple Negative (TN) [7]. Subclassification is necessary to better identify molecular-based therapeutic targets, select biomarkers, discover new drugs, and design clinical trials that will enable alignment of TNBC patients to appropriate targeted therapies. Cost and complexity issues can render gene expression profiling impractical as a routine hospital diagnostic tool, while immunohistochemical (IHC) marker detection is feasible for the majority of institutions. Nielsen et al (2004) [8] and Carey et al (2006) [9] showed that detection of epidermal growth factor receptor (EGFR) and/or cytokeratin 5/6 (CK5/6) expression by IHC staining can accurately identify the basal-like tumors among cohorts profiled by expression microarray with 100% specificity and 76% sensitivity [6]. This ‘‘fivemarker panel’’ ER-PR-HER2–EGFR-CK5/6 allows subclassification of TNBCs as basal-like (or Core Basal, CB) when EGFR and/or CK5/6 are positive or five negative (5NP) if all markers are negative. In our current study, we have correlated the 5 marker (ER-PR-HER2EGFR-CK5/6) IHC expression profiles with microRNA (miRNA) expression and generated a four-miRNA prognostic signature that stratifies with high specificity CB and 5NP by overall survival. We also have correlated the miRNA signature with prognosis and survival following treatment with different chemotherapy regimens. MiRNAs are 19–25 nucleotide, non-coding RNAs that reduce the abundance and translational efficiency of mRNAs and play a major role in regulatory networks,

www.impactjournals.com/oncotarget

RESULTS Definition of triple negative breast cancer biological subtypes by immunohistochemistry Two TMAs comprehensive of the TNBC cohort profiled for miRNA expression [17] were evaluated by immunohistochemical analysis to assess the expression of ER, PR, HER2, EGFR, and CK5/6. FISH analysis for the HER2 gene was also performed, with no gene amplification observed in all the tested cases.Cases were categorized based on their IHC profiles into two subclasses: (I) triple-negative cancers (i.e. ER-PR-HER2 negative) expressing EGFR and/or CK5/6, here referred to as CB, the so called “basal-like” as defined by mRNA expression analysis; and (II) cancers negative for the five markers, referred to as the 5NP subclass, triple negative cancers that express neither EGFR nor CK5/6, or “non basal” if considering the definition by mRNA expression.

1175

Oncotarget

Table 1. Clinical and demographic characteristics of the TNBC cohort. Entire cohort (n=173)

Characteristics

Number of cases Race Caucasian African American Other Menopause status Pre-menopausal Post-menopausal Unknown Grade I II III Unknown LN metastases Positive Negative Unknown Age at diagnosis =51 Death No Yes *Recurrence No Yes *Type of 1st recurrence In situ Local/Regional Distant Type unknown Chemotherapy None Anthracycline, (primarily AC, FEC) Anthracycline + taxanes (AC+taxol+taxotere) Non anthracycline, no taxanes (CMF) Taxane(s) + non anthracycline drugs Not Available info (NA)

IHC (n=160)

miRNA signature (n=160)

CB 92 (57.5%)

5NP 68 (42.5%)

High Risk 80 (50%)

Low Risk 80 (50%)

153 16 4

82 (89.1%) 7 (7.6%) 3 (3.3%)

61 (89.7%) 7 (10.3%) 0 (0%)

71 (88.6%) 7 (8.8%) 2 (2.6%)

72 (90%) 7 (8.8%) 1 (1.2%)

64 103 6

35 (38%) 54 (58.7%) 3 (3.3%)

21 (31%) 46 (67.6%) 1 (1.4%)

31 (38.8%) 47 (58.6%) 2 (2.6%)

25 (31.2%) 53 (66.2%) 2 (2.6%)

2 15 150 6

1 (1.2%) 6 (6.5%) 83 (89.1%) 2 (2.2%)

0 (0%) 8 (11.8%) 59 (88.2%) 1 (1.4%)

1 (1.2%) 9 (11.2%) 67 (83.8%) 3 (3.8%)

0 (0%) 5 (6.2%) 75 (93.8%) 0 (0%)

62 102 9

36 (39.1%) 50 (54.4%) 6 (6.5%)

22 (32.4%) 43 (63.2%) 3 (4.4%)

28 (35%) 45 (56.2%) 7 (8.8%)

30 (37.4%) 48 (60%) 2 (2.6%)

34 52 87

17 (18.5%) 30 (32.6%) 45 (48.9%)

14 (20.6%) 13 (19.2%) 41 (60.2%)

16 (20%) 27 (33.8%) 37 (46.2%)

15 (18.7%) 16 (20%) 49 (61.3%)

114 59

57 (62%) 35 (38%)

48 (70.6%) 20 (29.4%)

43 (53.7%) 37 (46.3%)

62 (78.7%) 18 (21.3%)

126 47

63 (68.5%) 29 (31.5%)

53 (77.9%) 15 (22.1%)

53 (66.2%) 27 (33.8%)

63 (78.7%) 17 (21.3%)

1 3 35 8

1 (1.2%) 2 (2.2%) 23 (25%) 3 (3.3%)

0 (0%) 1 (1.4%) 12 (17.8%) 2 (2.9%)

1 (1.2%) 2 (2.5%) 21 (26.3%) 3 (3.8%)

0 (0%) 1 (1.2%) 14 (17.5%) 2 (2.6%)

25

11 (12%)

11 (16.2%)

10 (12.5%)

12 (15%)

32

13 (14.1%)

16 (23.5%)

12 (15%)

17 (21.3%)

36

25 (27.1%)

9 (13.3%)

20 (25%)

14 (15.5%)

9

6 (6.5%)

2 (2.9%)

4 (5%)

4 (5%)

5

3 (3.3%)

2 (2.9%)

4 (5%)

1 (1.2%)

66

34 (37%)

28 (41.2%)

30 (37.5%)

32 (40%)

Tab. Abbreviations: CB (Core Basal); 5NP (five negative phenotype); AC, doxorubicin and cyclophosphamide; CMF, cyclophosphamide, methotrezate, and fluorouracil; FEC fluorouracil, epirubicin, cyclophosphamide; FAC, fluorouracil, doxorubicin, and cyclophosphamide. Percentages are referred within the subtype. 1= anthracycline containing regimen

www.impactjournals.com/oncotarget

1176

Oncotarget

Of the 160 TNBCs considered, 82 were negative for EGFR (51%), and 131 for CK5/6 (82%). A total of 92 cases (57.5%) were classified as CB, and 68 (42.5%) as 5NP. The clinico-pathological characteristics of the entire TNBC cohort, as well as of the two subclasses, are summarized in Table 1.

CI=1.04-2.92; p=0.03) (Figure 1B). Upregulation of miR-155 (HR=0.73; 95% CI=0.57–0.92) and of miR-493 (HR=0.88; 95% CI=0.72–0.99) correlated with better patient outcome so were defined as “protective”; downregulation of miR-30e (HR=1.08; 95% CI=1.03–1.79) and of miR-27a (HR=1.09; 95% CI=1.03–1.79, correlated with a worse outcome so were defined as “risk”-associated (Supplementary Table 2).Cox proportional hazards models were applied to find, significant associations of the four deregulated miRNAs with CB and 5NP patient outcomes. All tumors were classified into high- or low-risk groups according to their risk-score (see Materials and Methods). The Kaplan-Meier overall survival (OS) graph, according to the combined four deregulated miRNAs is shown in Figure 1C. The median OS for the high vs low risk miRNA signature were 75.5 vs 82 months (HR=2.46; 95% CI=1.43-4.12; p=0.001), indicating a significant association between expression of the miRNAs and OS.

miRNA expression profiles of the CB and 5NP subclasses identifies a diagnostic four miRNA signature To identify diagnostic miRNA signatures in TNBCs, miRNA expression profiles already analyzed for TNBC expression pattern (Supplementary Figure 1), were examined to find differently deregulated miRNAs among CB and 5NP tumors.Supervised clustering of the cohort based on the IHC results (Figure 1A) shows a signature of four miRNAs that performed best in differentiating between CB and 5NP cancers (Figure 1A and Supplementary Table 1): miR-155 (logFC 0.76; p=0.04), miR-493 (logFC 0.54; p=0.01), miR-30e (logFC -0.61; p=0.04), and miR-27a (logFC -0.80; p=0.01). This four miRNA signature displayed sensitivity 0.75 and specificity 0.56 (AUC=0.74) in subclassifying CB or 5NP.

Prognostic impact of TNBC subtype classification dependent on specific therapy regimens Since CB and 5NP subclasses have distinct OS as well high/low risk miRNA signature-based, we next analyzed the correlations among: subclasses, chemotherapy regimens and outcome.In the 107 patients that received chemotherapy, the CB group had significantly worse OS compared to the 5NP group (HR=2.46; 95% CI = 1.25-4.25; p=0.008) (Figure 2A). A

miRNA signature impacts survival of TNBCs Based on IHC subtyping, CB cancers exhibited a significantly worse outcome than 5NP (HR=1.76; 95%

FIGURE 1 (A, B, C): Identification of miRNAs that are differentially expressed in CB and 5NP breast cancers. A)

Heat map representing miRNA profiles of 160 tumor samples using complete linkage and Pearson correlation method as distance metrics. Orange identifies CB and Green 5NP tumors. Columns represent individual cancers; rows represent expression of miRNAs. Heat map colors represent relative miRNA expression as indicated in the blue to red key bar at the top. B) Overall survival of 160 TNBC tumors based on the status of the five IHC markers. C) COX proportional hazard model shows the overall survival based on the four miRNA high / low risk signature. www.impactjournals.com/oncotarget

1177

Oncotarget

similar trend was observed when stratifying the TNBCs for high and low risk by the miRNA signature (Figure 2B), the high-risk subgroup having a lower OS compared to the low-risk group (HR=1.98; 95% CI = 1.04-3.74; p=0.04). In patients not undergoing chemotherapy, both IHC and four miRNA signature subtyping significantly stratified patients in prognostic classes (Figures 2C and 2D). 5NP untreated patients showed a better outcome compared to the untreated CB group (HR=2.97; 95% CI=1-8.87; p=0.05). The four miRNA signature subclassification performed better in discriminating high risk patients in comparison to the IHC subtyping (HR=6.19; 95% CI = 2.45-32.16; p=0.001).If the group of patients for whom chemotherapy treatment information was not available (NA) are stratified according to IHC status there is a clear trend toward worse survival of the CB subtype (Supplementary Figure 2A); also the four miRNA predictor can even more clearly separate this cohort into high/low risk cancers (HR=4.07; 95% CI = 1.44-9.75; p=0.007) (Supplementary Figure 2B). TNBC patients are usually treated with a regimen of anthracycline or anthracycline plus taxanes, regardless of their proven intrinsic heterogeneity or IHC status. As shown in Table 1, within the entire TNBC cohort, four main chemotherapeutic regimens were administered: 1anthracycline containing regimen (primarily doxorubicin and cyclophosphamide [AC], fluorouracil, epirubicin, cyclophosphamide [FEC]); 2- anthracycline + taxanes

(AC+ taxol or taxotere); 3- non-anthracycline, no taxane containing regimen (cyclophosphamide, methotrexate, fluorouracil [CMF]); 4- taxane alone or in combination with non-anthracyline drugs (primarly taxol alone). Among the 92 CB TNBC patients, 11 received no chemotherapy, 38 were treated with anthracyclinebased chemotherapy (13 primarily with doxorubicin/ cyclophosphamide and 25 with doxorubicin, taxol, taxotere) and 9 were treated with non-anthracycline-based chemotherapy (6 with cyclophosphamide, methotrexate, fluorouracil and 3 taxane or taxanes). Among the 68 5NP TNBC patients, 11 received no adjuvant systemic therapy, 15 received anthracycline-based chemotherapy (6 primarily doxorubicin/cyclophosphamide and 9 doxorubicin, taxol, taxotere), and 2 received nonanthracycline-based chemotherapy (Table 1). We stratified the cohort into the CB and 5NP subclasses and then considered anthracycline containing regimen vs untreated patients (Figure 3A). Besides the not unexpected positive effect of the chemotherapy in both subclasses, it is observed that the untreated 5NP group showed longer life expectancy than the chemotherapy treated CB group (p=0.027). This finding underlines the much better prognosis of the 5NP vs the CB cancers regardless of chemotherapy, as well as the lower efficacy of this therapy for the 5NP group. The high/low risk miRNA signature overall survival shown in Figure 3B is comparable to the IHC-based (Figure 3A), where the

FIGURE 2 (A, B, C, D): Overall survival of chemotherapy-treated and untreated TNBC patients. A) IHC based overall survival of CB vs 5NP patients receiving chemotherapy; B) COX proportional hazard survival model of chemotherapy-treated patients stratified by high/low risk 4 miRNA signature C) Overall survival of IHC-stratified CB and 5NP untreated patients D) COX proportional hazard survival model of untreated patients based on 4 miRNA signature predictor www.impactjournals.com/oncotarget

1178

Oncotarget

anthracycline-containing regimen significantly influenced the prognosis of high-risk tumors (p